Thu Jul 29 12:00:45 PDT 2004
T0254

DUE 22 Aug 2004


Thu Jul 29 09:50 PDT 2004 Kevin Karplus

Observation:

The reason (at least a reason) try1 was so messed up was that the
alignment all-align.a2m.gz was not getting properly created before
try1 was run.  I moved try1 to no_align1, and am rerunning try1 after
creating all-align.a2m.gz properly.


Fri Jul 30 05:26:37 PDT 2004 Kevin Karplus

redone try1-opt looks pretty good, coming almost directly from the templates.
The helices need to pack a bit tighter against each other.


Sun Aug 15 08:48:48 PDT 2004	Sol Katzman

The template 1dpt is a homo-trimer, but there are only small
trimerization interfaces -- most sides of each monomer are exposed.

Beyond Kevin's suggestion about packing the helices, we might want to:
a) fix breaks at F60-G61 and H87-L88
b) get a little longer Hbonding between strands s1 || s2
c) get rid of the bogus strands/sheets Y100-Q105 and E48-V51

For try2, include most SheetConstraints and HelixConstraints from
try1, and increase the weights of hbond_geom_beta, break,
soft_clashes. Use only the constraints from t2k.dssp-ehl2, eliminating
the other predicted constraints.

Sun Aug 15 20:43:51 PDT 2004	Sol Katzman

It looks like try2 accomplished some of its goals:

a) breaks are improved somewhat. now the worst ones are at E63,P53,L88
b) we have better bonding for s1 || s2
c) strands E48-V51 eliminated, but Y100-Q105 still forms a short hairpin
   at the very C-terminus of the protein. I suppose this could be valid.

For a try3 polishing run, eliminate TryAllAlign, use read-pdb.under.
Use the try2 Helix and Sheet constraints (instead of try1). For the
first polish, leave the break (and other weights) unchanged.


Sun Aug 15 22:26:37 PDT 2004	Sol Katzman

fooey! caught an assertion check from undertaker. This is the same assertion
that we were seeing when running T0239 trys on the condor cluster (with
the old compiler). In the current situation, we were running on 'hori'
using the standard stuff from the Makefile.

undertaker: GeneticMethod.cc:59: static void GeneticMethod::set_cum_probs(): Assertion `sum>0' failed.

double fooey! it looks like my original try3.under file still had "try2"
for all the pointers. But what luck! Due to the crash, none of the try2
models were overwritten.

In hopes that a random seed will make this go away, I am going to resubmit
try3, just renaming it as try4. I am also going to rename the try3.under
file to try3.under.bad. I hope this is not all too confusing.

   try1 -- normal
   try2 -- normal
   try3 -- crashed, and had "try2" in it but did not overwrite try2 output.
           renamed file to try3.under.bad
   try4 -- a copy of what should have been the try3 run. will it crash?


Mon Aug 16 08:17:52 PDT 2004	Sol Katzman

Okay, try4 succeeded (no assertion). But breaks and clashes are
essentially unchanged from try2.

Maybe one more polishing run: try5 with constraints omitted, and
increase break (40->45) and dry5 (5->9) dry6.5 (6->10)


Mon Aug 16 14:39:26 PDT 2004	Sol Katzman

Aha, insight. At our weekly meeting, it was suggested that I look for
other multimer templates to see why the one face of the sheet is exposed.

As mentioned above, the 1dpt template trimer has very little contact
between monomers.

But the 1mfi and 1hfo templates also form trimers with the 3 sheets
(one from each monomer) facing inward. Furthermore, there are specific
interfaces corresponding to the T0254 hairpin T100-F107 (call these
strands s5,s6, and the T0254 wayward strand T43-A47 (call this s2b).
In 1mfi these are not actual strands, but in 1hfo they are. In either
case, the pattern is that in a trimer composed of chains :A, :B, :C

     s6:A ^v s5:A ^v s4:B || s3:B ^v s1:B || s2:B  (A-B interface)
            s2b:A ^v s2:C || s1:C ^v s3:C || s4:C  (A-C interface)

It is a rather beautiful structure. To see it:

   cd ~/casp6/T0254
   rasmol /projects/compbio/data/pdb/1hfo.pdb.gz
   
   script 1hfo.pretty.rasmol

For try6, I will reinstate the StrandConstraints to create s2b
and add the SheetConstraint for s6 ^v s5. Then I will trimerize
based on chains A,B,C of 1hfo and see how it fits together.

Mon Aug 16 17:24:32 PDT 2004	Sol Katzman

try6 failed with the undertaker assertion.

try6.under and .costfcn copied to try7 and rerun.